denote the waiting time until first entry into a two-year and four-year institutions,
respectively. In this case equation (1) above is replaced by
k2y
S(k2y,k4y∣z,θ2y β4y) = exp- θ2y∑ exp(rr2y + (β2y)‘z)
r=1
(1')
k4y
-∑ expYy + (βr4y)‘z)]
r=1
where the parameters in (1N) have interpretations analogous to those described above.
Now, however, we have a quadruplet of location parameters (θ2y, θ4y, θg, θd) with
distribution function G( θ2y,θ4y, θg, θd). Again, we assume a mass point structure where
each of the N quadruplets occurs in the population with proportion pn:
N
∑ pn = 1.
n=1
When focusing on those first entering four-year (two-year) institutions, those first
entering two-year (four-year) institutions will subsequently be censored in the likelihood
function.5
4. Data
5The likelihood function for the CR2 model can be derived in a manner similar to that for the DCR model
(see Appendix B). For the sake of brevity, however, we omit its derivation in this paper.
12